JP2023553676A - Delivering explainable machine learning model results using a distributed ledger - Google Patents

Abstract

receiving input data for a machine learning (ML) model; processing the input data using the ML model to produce an initial result; and adding a first block to a distributed ledger. , the block includes input data, an initial result, an ML model data structure, and a link to training data for the ML model, the training data is present in the preceding distributed ledger block, and includes the initial result. Providing reproducible machine learning model results by providing output.

Description

The present disclosure generally relates to providing explainable machine learning model results. The present disclosure particularly relates to providing explainable machine learning model results as the machine learning model matures.

Machines are becoming highly intelligent. In a smart way, these machines not only serve humans through intangible interfaces (cognitive devices that can communicate with humans in natural language), but also through tangible interfaces (robots or other tangible interfaces). serve humans. Cognitive interfaces have machine intelligence, have the ability to process input data, and can also obtain additional information from resources for better processing by communicating with other devices.

Modern complex artificial intelligence (AI) techniques, such as deep learning and genetic algorithms, are opaque in nature. AI systems actively rely on the training corpus to learn from it and make decisions in a more natural manner. This constant learning improves the performance of the AI system daily based on more relevant learning and history. Sometimes, due to the complexity of machine learning (ML) models, AI systems learn undesirable classification pathways. These undesirable classification pathways lead to interpretability problems of AI results.

In a black box type ML model, stakeholders cannot explain why the AI reached a particular decision. Therefore, there is little data to justify the information and results produced by AI systems. Explainable AI (XAI) refers to methods and techniques in the application of artificial intelligence techniques (AI) in which the results of a model can be explained and understood by human experts. XAI provides explainability and interpretability of ML model results. XAI models generate a lot of metadata to provide the necessary explanations and evidence that can be manually verified, giving end users a desired level of confidence in the ML model results. XAI gives a score that describes how much each input element contributed to the final result of the model prediction. This supports the concept of ``Safe AI,'' which allows humans to know what is being decided inside the AI's ML model. ML models that do not provide such explanations may not be applicable to business-critical decisions.

In a typical ML model, the primary contributor that produces an output or set of outputs is the corpus used to train the model. As the model processes the data, it generates results and appends metadata mapper objects associated with the results to the training corpus. These metadata mapper objects contain information derived during generation of the first set of outputs. As the ML model learns more, it becomes more mature due to the backfilling of metadata into the training corpus that was generated as a byproduct of current and previous runs.

The daily maturation of the ML model affects the output produced by the ML model derived using the ML model's evolving index. As a result, the same set of input data may produce different outputs at different times, based on the maturation of the ML model. Furthermore, due to changes in the maturing model, there may be no means to retrospectively verify or explain the output from the ML model after a period of time.

For example, once an ML model matures and its state changes, there is no mechanism that can verify and explain how previous results from that model satisfy anti-discrimination provisions. . The dynamic state of an ML model changes based on the evolution of the model's training corpus. Therefore, if ML model decisions are not saved in a proper manner, there is no way to audit them over time. Furthermore, there is no method that can perform multiple validations over a significant period of time. Although there are mechanisms to store intermediate results of ML models as part of test validation evidence, there are multiple ML models that use a common training corpus, and each ML model stores multiple intermediate result data. In this case, the size of the model results and training corpus set becomes unmanageable.

A summary is presented below to provide a basic understanding of one or more embodiments of the disclosure. This summary is not intended to identify key or critical elements or to delineate any scope of particular embodiments or the claims. Its sole purpose is to present concepts in a simplified form as a prelude to the more detailed description that is presented later. In one or more embodiments described herein, a device, system, computer-implemented method, apparatus, and/or computer program product provides explainable and reproducible machine learning model results. becomes possible.

Aspects of the invention include receiving input data for a machine learning model, processing the input data using the machine learning model to produce an initial result, and adding a first block to a distributed ledger. adding, the block includes input data, an initial result, an ML model data structure, and a link to training data for the ML model, the training data being present in a previous distributed ledger block; Disclosed are methods, systems, and computer-readable media associated with providing reproducible machine learning model results by providing output that includes initial results. This method provides distributed and immutable tracking of ML model transformations and allows auditing and validation of past ML model results.

Aspects of the invention include receiving input data for a machine learning model, processing the input data using the machine learning model to produce an initial result, and adding a first block to a distributed ledger. adding, the block includes input data, an initial result, an ML model data structure, and a link to training data for the ML model, the training data being present in a previous distributed ledger block; receiving a result validation request for input data; retrieving the input data and an ML model data structure from the first block; and processing the input data using the ML model data structure to produce a validation result. Disclosed are methods, systems, and computer-readable media associated with providing reproducible machine learning model results by providing output that includes initial results and validation results. This method provides distributed and immutable tracking of ML model transformations and allows auditing and validation of past ML model results.

Aspects of the invention include receiving input data for a machine learning model, processing the input data using the machine learning model to produce an initial result, and generating an ML model description associated with the initial result. determining and adding a first block to the distributed ledger, the block including input data, initial results, an ML model description, an ML model data structure, and a link to training data for the ML model; associated with providing reproducible machine learning model results by appending the training data present in the preceding distributed ledger block and providing an output that includes the initial results and the ML model description. , methods, systems, and computer-readable media are disclosed. This method provides distributed and immutable tracking of ML model transformations and allows for explanation, auditing, and validation of past ML model results.

Aspects of the invention include receiving input data for a machine learning model, processing the input data using the machine learning model to produce an initial result, and adding a first block to a distributed ledger. adding, the block includes input data, an initial result, an ML model data structure, and a link to training data for the ML model, the training data being present in a previous distributed ledger block; providing an output containing the initial results; updating the ML model using the input data and results; generating a new ML model version; and adding a second block to the distributed ledger. and adding a second block containing labels associated with the new ML model version, input data, initial results, ML model data structure, and a link to training data for the previous version ML model. Discloses methods, systems, and computer-readable media associated with providing reproducible machine learning model results. This method provides distributed and immutable tracking of ML model transformations and allows auditing and validation of past ML model results.

These and other objects, features, and advantages of the present disclosure will become more apparent through a more detailed description of some embodiments of the present disclosure in the accompanying drawings. In embodiments of the present disclosure, like numbers generally refer to like components.

1 is a schematic diagram of a computing environment, according to an embodiment of the invention; FIG. 2 is a flowchart depicting an operable sequence according to an embodiment of the invention. 2 is a schematic diagram of data flow according to an embodiment of the invention; FIG. 1 is a diagram of a cloud computing environment, according to an embodiment of the invention. FIG. FIG. 2 is a diagram of abstract model layers, according to an embodiment of the invention.

Some embodiments will now be described in more detail with reference to the accompanying drawings, in which embodiments of the disclosure are illustrated. However, this disclosure may be implemented in a variety of ways and therefore should not be construed as limited to the embodiments disclosed herein.

The disclosed embodiments address issues of explainability and reproducibility of machine learning model results. Embodiments enable regeneration of results produced during previous implementations of the model and provide a mechanism to validate time-based ML model output. The disclosed embodiments include a blockchain ledger as one of the intermediate components. After receiving input data from various entities, such as static and dynamic computing platforms associated with the ML model, disclosed embodiments create blocks on the blockchain ledger and update existing blocks in the ledger. Generate appropriate links with blocks. The disclosed embodiments include a blockchain ledger-backed producer-consumer architecture, where the entities involved in ML model computation are blockchain stakeholders.

At the time of computing the ML model output for a given set of problem and input information, the producer engine in the implementation creates blocks on the distributed ledger and updates the ML model with PRODUCTION signals. In response, the implemented consumer daemon mechanism in the ML model collects the associated training corpus data or previous ML model versions and node weights from the ledger so that the trained ML model Applied to input data. Consumers have a clear understanding of the individual stakeholders involved in the calculation. A consumer daemon associated with the ledger and ML model writes intermediate results to the ledger. Many ML models implement a path-oriented architecture in which multiple intermediate data structures are generated and posted to a ledger.

In embodiments, the ML model creates blocks on the ledger architecture for intermediate results, tags all participants - stakeholders, and creates blocks with data associated with the current ML version and the current input. Save the output. Input stakeholder data is written directly to the ledger for consumer processing, while training corpus data is collected from existing blocks created during previous runs of the ML model. While fetching from the training corpus, the ML model collects flags or data location links (tagging data) along with the actual training corpus dataset. The disclosed embodiments create a data structure for new ML model versions, and old links within objects are preserved while the ML model history is updated. Retaining a link to this old training corpus tagging provides backward traceability to the tagging corpus and allows for re-creation of the ML model if needed or requested.

If the ML model needs to be validated after the maturity index changes, the disclosed embodiments utilize a reproducibility special FLAG that activates the reproducibility of the cognitive system in the consumer instance of the ML model. . The consumer daemon in the ML model selects training corpus data tagged on the ledger and searches for associated time-based stakeholder input data accordingly. Once this input data has been collected, the disclosed embodiments fetch the training corpus by exploring a ledger tagging data structure and tracing links based on the time of previous run executions. . Consumer/user and producer daemons refer to the utility program portion of the disclosed method. These utility program parts perform the disclosed method activities for consumers/users and producers, respectively. Such steps include generating/signing ledger entries associated with inputs and outputs, generating/signing blocks containing associated ledger entries, requesting output validation, and performing validation/auditing. searching for and utilizing the necessary ML model components to

The ML model results, the audit trail in the ledger, and the ML model description can be verified using the exact same training corpus and ML model data structure that was used to output the original results. At each confidence level of reporting/output from the data model, the disclosed method invokes an explainability module. This module creates blocks on the blockchain and updates the current state of the model with an associated description. After providing a decision/output, all audit trails are created for that decision, including the data sources used, the data structure of the ML model used, the ML model training corpus, and any trust explanations along the way. . The disclosed method uses smart contracts or manual controls to course correct the model and creates an audit trail of such adjustments.

In embodiments, one or more components of the system may utilize hardware and/or software to solve problems that are highly technical in nature (e.g., train trained machine learning models). adding blocks to a distributed ledger, where the blocks include input data, machine learning model data structures, intermediate model results, model descriptions, and model training within the distributed ledger. including and adding tags to corpus locations, providing model processing results, etc.). These solutions are not abstract and can be implemented as a series of mental acts by humans, e.g. due to the processing power required to facilitate reproducible and explainable machine learning model results. It is not possible. Furthermore, some of the processing performed may be performed by specialized computers to perform defined tasks for explainable and reproducible machine learning model results. For example, specialized computers can be utilized to perform tasks involving reproducible and explainable machine learning model results that are accessible across distributed ledgers, such as by participating stakeholders.

In embodiments, a method for AI data governance and explainability receives input data. The input data relates to the intended use of the trained machine learning model, such as pass-fail classification based on the model's current node weightings and processing of the input data. Users enter data as digitally signed ledger entries that link users, data, and producers. The ledger entry includes the user's digital signature and the user's and producer's public keys. In embodiments, the producer receives input data from a user. In this embodiment, the producer creates and signs ledger entries.

The ledger entry includes the input data, the user's public key, the producer's public key, and the producer's digital signature. Ledger entries may include identifying information data about users, producers, or both users and producers. The method processes input data using the trained model and produces a result, such as a pass or fail classification of the input data. As an example, a simple linear regression model receives a given dataset as input and produces an output that includes attributes about the input records. Input data and output from the model are captured together as submitted ledger entries that are authenticated by the consensus of blockchain users and later added to new blocks on the blockchain. Consensus-based authentication of submitted ledger entries using digital signatures and available public keys prior to their addition to the ledger and blockchain preserves the integrity of the ledger.

After determining the results for the input data, the method creates new blocks for a distributed ledger, such as a blockchain. A blockchain may be a private distributed ledger in which the identities of participants are mutually known, or it may be a more public distributed ledger in which the identities of participants are hidden. Blockchains may be created using the open source HYPERLEDGER or ETHEREUM blockchain description platforms, or other blockchain platforms. (Note: The terms "HYPERLEDGER" and "ETHEREUM" may be subject to trademark rights in various jurisdictions around the world and are used herein solely for products or services appropriately named by the trademarks. , used as a reference to such trademark rights to the extent they may exist.)

Each participant has an identification information (ID), a public-private key pair, and a digital signature. The participants' digital signatures can be based on public-private cryptographic key pairs. The id may include the public key of the participant. A combination of identity and digital signature can be used to authenticate any transaction. In embodiments, each transaction carries a transaction id, an originator id, and the originator's digital signature encoded using the originator's private key and the transaction id. The originator id, or originator public key, can then be used to decrypt the transaction id from the digital signature that authenticates the origin of the transaction.

By using a public-private key pair, entries to the ledger that are created and signed using digital signatures, such as input data ledger entries signed by users and ML model result entries signed by producers, can be created and signed using digital signatures. , can be verified by those involved. Entries can be verified without information of the originator's identity. In embodiments, the size of each block is predetermined. When a predetermined size limit is reached, a new block is created that incorporates the hash of the previous block. In this embodiment, the method creates a new block with each result during processing. The method creates a block for each intermediate result, as well as one block for the final result of the processing.

In one embodiment, the method creates a ledger entry for each intermediate result and creates one new block after determining the final result. The created block contains the user's identity, intermediate and/or final results, and the version or data structure of the machine learning model used to achieve the result, such as the current matrix of node weights for the model. is included. A block also includes one or more tags. Each tag gives information about the location of the data used as the training corpus with respect to the current version of the machine learning model. The tag points to the previous block of the distributed ledger where the training corpus data was stored. A block is one or more entities that link stakeholders of the current outcome, such as a user entity that provides input data and a producer entity that holds the ML model and utilizes the decision results for decisions involving the user entity. further includes ledger entries for . Ledger entries include smart contract entries. Smart contracts allow users and producers to each use appropriate versions of machine learning models to verify and reproduce results. In embodiments, the smart contract further enables the recreation or retraining of the machine learning model using tagged training corpus data retrieved from previous blocks of the ledger. A smart contract may include conditions that must be satisfied to generate validation of prior results. Such conditions include agreement from each of the producers and users to generate verification results.

The method generates an XAI explanation, in some embodiments, at least for the final result, and in some embodiments, for each intermediate ML model result as well as the final result. The method generates an XAI description using an XAI methodology. The XAI description provides information about the relative weighting of each piece of input data on the decision result, thereby explaining the impact that each piece of input data has on the result. For example, for a set of input data containing n data points, an indicates whether each data point detected impairs the decision. The XAI description further includes the relative weighting of the set of n data points, from highest weighted and most influential to lowest weighted and least influential. In this embodiment, the method further includes XAI descriptive data in the blocks created for the results.

In embodiments, each created block includes hash values derived from all previous blocks of the ledger. In this embodiment, the method uses a hash function such as SHA-256 to generate a hash value from the previous block. In this embodiment, blocks are signed using the producer's private key-based digital signature. Consensus verification of blocks occurs through users and other participants who verify digital signatures using the producer's public keys.

In embodiments, the method provides the determined results as output to the user and producer. The method may further provide users and producers with an indication of ledger entries and blocks that store outcome-determining transactions.

Over time and through the generation of a large number of results, data structures such as ML model node weights evolve. The method expands the training corpus for the model by adding each combination of input data and decision results. As the training corpus expands, the model data structure evolves and new versions of the model are created. In embodiments, the method stores each new version along with the blocks created, including a ledger entry for the model version number and associated data structure. The entry further includes a tag to the expanded training corpus used to derive the new data structure.

After the evolution or maturation of the ML model data structure, the method adds appropriate ledger entries and blocks that store the training corpus and evolution of the ML model data structure to the blockchain. Following this addition, users, producers, or other stakeholders may attempt to reproduce and verify previous results. Because model versions and data structures have evolved, replicating previous results requires using the model version originally used rather than the current version. Because stakeholders are associated with the desired results through associated ledger entries, either users, producers, or other relevant stakeholders can request verification and reproduction of the results. The requesting stakeholder submits a request signed using his digital signature. In embodiments, the method verifies the digital signature of the request using the stakeholder's public key.

After verifying the signature, the method retrieves data regarding the verification request from blocks of the blockchain. The method retrieves relevant user input data, model versions and data structures, training corpus links, original intermediate and final results, and associated XAI descriptions from appropriate blocks to complete the validation request. Completing a validation request involves reprocessing input data using ML model data structures to produce intermediate and final validation results. The method then compares the original and validation intermediate and final results. In embodiments, the method further compares the original and verified XAI descriptions. Since the method processes the original input data using the original ML version and data structure, the original and validation results and XAI descriptions must be identical. The method provides the original and the results of the verification to the verification requester.

In embodiments, the method processes the retrieved input data using a model that conforms to the retrieved model data structure and verifies that the retrieved ML models determine the same results and the same explanations. . In embodiments, the method uses the tags to further search the training corpus data. In this embodiment, the method uses the retrieved training corpus to create an ML model to verify that the training corpus yields a retrieved ML model data structure associated with the retrieved model version. Re-create. The method provides the output of the requested result validation to the validation requester. In embodiments, the method creates a ledger entry that documents the request and associated output for the verification and the results of the verification (results are repeated or failed verification results are not repeated). In embodiments, the method provides notification of the request for validation and output of the requested validation to all stakeholders associated with the results. In this embodiment, all method steps are stored in associated ledger entries that are incorporated into the next new block.

Disclosed embodiments enable auditing, validation, and replication of ML model results as a means of satisfying regulatory requirements related to the use of ML models to make pass-fail decisions or for other uses. It becomes possible. The disclosed embodiments provide a replica of the ML model used in determining initial results through the ML model data structure and use the original data corpus to verify training. and provide an XAI explanation of model decisions to satisfy regulatory requirements regarding ML model transparency.

In embodiments, the producer may manually intervene in the determination of model results, for example changing the model determination from fail to pass. Such intervention requires a suitable digital signature of the intervention from the producer. In this embodiment, the method uses the producer's public key to verify the digital signature, make the requested changes, and track manual intervention and its effect on the model's data structure. The method stores this effect in a ledger entry as a new version of the model, documenting the manual intervention and modification of the results. The method creates a new block that captures ledger entries associated with the intervention. The method adds tags to the manual results as part of the training corpus for this version of the model, as well as for all subsequent versions. In embodiments, the method adds tags to ledger entries associated with the intervention.

In embodiments, the method retrains the ML model using the manual intervention results and related input data to generate a modified version of the ML model and an associated modified data structure. The method labels the new ML version, includes effectively labeled input data (input data and manual intervention results), and stores the new version, data structure, and training corpus. The method generates new ledger entries and associated new blocks, including new versions, input data, manual intervention result labels, version labels, manual intervention indications, and the like. The producer signs a new block and submits it for signature-based consensus validation.

In embodiments, the method includes one or more application programs for providing input data, submitting audit or validation requests, and receiving output data from the model. Contains an interface (API). The API links stakeholders to the ledger and the state machine built using the model and training corpus. The API provides a pathway for user input data and producer direct/manual intervention data to the state machine. The API provides a means for the generation of new ledger entries and new blocks that store model results and tag model training corpus data locations in previous blocks. The API further enables the generation of ledger entry smart contracts that link stakeholders to their common outcomes.

FIG. 1 is a schematic diagram of example network resources associated with implementing the disclosed invention. The invention may be practiced with any processor of the disclosed elements that processes a stream of instructions. As shown in the figures, networked client devices 110 wirelessly connect to server subsystem 102 . Client device 104 wirelessly connects to server subsystem 102 via network 114 . Client devices 104 and 110 include an application program interface (not shown) associated with the disclosed embodiments, along with sufficient computational resources (processor, memory, network communication hardware) to execute the program. .

In embodiments, client devices 104 and 110 include users and producers associated with the disclosed systems and methods. Users and producers connect through network 114 to provide input data, use ML models to generate results, create and validate blocks, and add them to the blockchain.

As shown in FIG. 1, server subsystem 102 includes server computer 150. FIG. 1 is a block diagram of components of a server computer 150 in a networked computer system 1000, according to an embodiment of the invention. It is to be understood that FIG. 1 merely provides an illustration of one implementation and does not imply any limitations as to the environment in which various implementations may be implemented. Many modifications can be made to the depicted environment.

Server computer 150 includes a processor 154, memory 158, persistent storage 170, communications unit 152, input/output (I/O) interface 156, and communications fabric 140. Communication fabric 140 provides communication between cache 162, memory 158, persistent storage 170, communication unit 152, and input/output (I/O) interface 156. Communications fabric 140 is configured to pass data and/or control information between processors (such as microprocessors, communications and network processors), system memory, peripheral devices, and any other hardware components within the system. It can be implemented on any architecture designed in For example, communications fabric 140 may be implemented using one or more buses.

Memory 158 and persistent storage 170 are computer readable storage media. In this embodiment, memory 158 includes random access memory (RAM) 160. In general, memory 158 may include any suitable volatile or nonvolatile computer-readable storage medium. Cache 162 is a high speed memory that improves the performance of processor 154 by retaining recently accessed data from memory 158, as well as more recently accessed data.

Program instructions and data used to implement embodiments of the present invention, such as data governance and Stored in persistent storage 170 for execution and/or access. In this embodiment, persistent storage 170 includes a magnetic hard disk drive. As an alternative to, or in addition to, a magnetic hard disk drive, persistent storage 170 may include a solid state hard drive, a semiconductor storage device, a read only memory (ROM), an erasable programmable read only memory (EPROM), a flash - May include memory or any other computer readable storage medium capable of storing program instructions or digital information.

The media used by persistent storage 170 may also be removable. For example, a removable hard drive may be used for persistent storage 170. Other examples include optical and magnetic disks, thumb drives, and smart cards that are inserted into drives for transfer to another computer-readable storage medium that is also part of persistent storage 170.

Communication unit 152 is provided for communication with other data processing systems or devices, including resources of client computing devices 104 and 110 in these examples. In these examples, communication unit 152 includes one or more network interface cards. Communication unit 152 may provide communication through the use of physical communication links, wireless communication links, or both. Software distribution programs, as well as other programs and data used to implement the invention, may be downloaded to persistent storage 170 of server computer 150 through communication unit 152.

I/O interface 156 allows data input and output to and from other devices that may be connected to server computer 150. For example, I/O interface 156 provides a connection to external device 190, such as a keyboard, keypad, touch screen, microphone, digital camera, or any other suitable input device or combination thereof. External devices 190 may also include portable computer-readable storage media, such as, for example, thumb drives, portable optical or magnetic disks, and memory cards. Software and data used to implement embodiments of the invention, such as data governance and XAI programs 175 on server computer 150, can be stored on such portable computer-readable storage media. and can be loaded into persistent storage 170 via I/O interface 156. I/O interface 156 also connects to display 180.

Display 180 provides a mechanism for displaying data to a user and may be, for example, a computer monitor. Display 180 can also function as a touch screen, such as a tablet computer display.

FIG. 2 is a flowchart 200 illustrating example activities associated with implementing the present disclosure. After starting the program, the XAI data governance program 175 receives input data at block 210. Input data may be passed to the program through an API from stakeholders such as users and producers. The input data may be digitally signed by the data provider and passed as a ledger entry containing public keys for all stakeholders up to the relevant transaction.

At block 220, the method of the XAI data governance program 175 processes input data from the ledger entries using the trained ML model. ML models include data structures derived during training of the model using a training corpus of data. ML models include model version indicators. Processing the input data results in one or more results associated with the input data, such as one or more pass-fail results associated with the underlying transaction between the user and the producer. Results include an XAI description of the final results and any intermediate results generated during processing of the input data by the ML model. XAI explanations allow humans to interpret and explain the results given as output by the model. In embodiments, the XAI results include relative weightings of the input data used in determining the results.

At block 230, the method of the XAI data governance program 175 adds the ledger entry to the distributed ledger. The ledger entry includes the identity of the stakeholder up to the processing event, as well as smart contract terms that describe the conditions necessary for the stakeholder to access the results or request auditing or other verification of the results. The additional ledger entries include the original ledger entry containing the input data, as well as the entry containing the ML model version indicator, and the ML model data structure and tags are based on the training corpus on which the ML model is based and the training corpus version indicator. Identify the block of the distributed ledger blockchain where the data resides. The method further generates new blocks for the blockchain. A new block contains a new ledger entry as well as a hash of the previous block in the blockchain. The producer signs the new block using a digital signature derived using the producer's private key. The method adds a new block after consensus validation of the block by stakeholders. The stakeholder verifies the new block using the producer's public key and verifies the digital signature of the new block.

The method adds input data and associated results to a training corpus of an ML model and retrains the ML model using the corrected training corpus. The method identifies new versions of ML models and new versions of ML model data structures, and tags additional training corpus data as part of the overall training corpus.

At block 240, the method of the XAI data governance program 175 provides the results of the processing, including direct pass-fail type results, in addition to the XAI descriptions of those results. The method may provide results to any combination of users, producers, and other stakeholders identified within the original ledger entry.

At block 250, the method of XAI data governance program 175 receives a request for validation or auditing associated with prior results or input data. The request satisfies the requirements for requiring validation as described in the ledger entry smart contract for input data. The method verifies the signature of the verification requestor and proceeds to complete the verification or audit request.

At block 260, the method of the XAI data governance program 175 stores the input data, original results, XAI description, ML model version and associated ML data structures, and training dataset location tags for the ML model version. Search from related ledger entries or previously added block entries. At block 270, the method processes the input data using the ML version based on the retrieved data structures to determine new intermediate and final results and XAI descriptions for these new results.

At block 280, the method provides the results of the validation, including the original results and XAI description and the new results and XAI description, to the requestor of the validation and possibly all other stakeholders up to the transaction. Providing both the original and new results allows comparison of the two sets of results by the verification requester. Providing XAI descriptive results for original and new results allows for auditing of the weighting of input data used to determine results.

FIG. 3 is a schematic diagram 300 of data flow according to an embodiment of the invention. As shown in the figure, input data flows from user 302 to producer 306 to distributed ledger state machine 320 as digitally signed ledger entries 325. The method of XAI data governance program 175 of FIG. 1 passes ledger entries 325 to ML model 330. The ML model 330 processes the ledger entry 325 input data and returns a result (such as pass or fail) that includes a classification for the input data, as well as XAI explanations for all intermediate and final ML model results. The output results and the links between the stakeholders (users 302, producers 306, etc.) associated with the outputs are documented in a ledger entry 327 along with an XAI description of the results. Smart contract ledger entries 329 describe the conditions for stakeholders to access results and explanations, as well as to request verification or auditing of results by stakeholders. ML model 330 provides results and XAI descriptions as output to transaction stakeholders, including users 302 and producers 306, through an API.

The XAI data governance program 175 method creates a new block 340 for the distributed ledger blockchain and signs the new block using the producer's digital signature. The new block includes the original ledger entry, the resulting ledger entry, the ML model data structure and version label, and the smart contract ledger entry. New blocks are added to the blockchain and verified by blockchain stakeholder consensus using the producer's public key.

The XAI data governance program 175 method appends input data and associated results to a training corpus and uses the appended training corpus dataset to train a new model version with revised data structures. /generate. Ledger entries are generated indicating the new version, revised ML model data structure, and additional training set tags, and yet another new block 350 is generated to store these ledger entries. Model 330 reads block ledger entries 325, 327, 329 captured in blocks 340 and 350 as required for validation of prior results.

Implementations of the disclosed embodiments may utilize local or networked computational resources. In embodiments, local resources connect to edge clouds or cloud resources to take advantage of additional computational resources available through such connections.

Although this disclosure includes detailed discussion of cloud computing, it is to be understood that implementations of the teachings described herein are not limited to cloud computing environments. Rather, embodiments of the invention may be implemented in conjunction with any other type of computing environment now known or later developed.

Cloud computing provides convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, network bandwidth, servers, processing, memory, storage, applications, virtual machines, and services). - A model of service delivery to enable access that can be rapidly provisioned and released with minimal administrative effort or interaction with the provider of the service. The cloud model can include at least five features, at least three service models, and at least four deployment models.

Features are as follows:

On-demand self-service: Cloud consumers can unilaterally provision computing capabilities, such as server time and network storage, automatically and as needed without requiring human interaction with the service provider. Can be done.

Broad network access: Functionality is available over the network and accessed through standard mechanisms that facilitate use by disparate thin or thick client platforms (e.g., mobile phones, laptops, and PDAs). Ru.

Pooling of resources: A provider's computing resources are pooled to serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically allocated based on demand. and reallocated. There is a sense of location independence in that consumers generally have no control or information about the exact location of the resources provided, but at a higher level of abstraction (e.g., country, state, or data center) can sometimes be identified.

Rapid scalability: Features can be provisioned quickly and scalably, and in some cases automatically, scaled out quickly, released quickly, and scaled in quickly. To consumers, the functionality available for provisioning often appears unlimited and can be purchased in any quantity at any time.

Service metering: Cloud systems automatically measure resource usage by leveraging metering capabilities at some level of abstraction appropriate to the type of service (e.g., storage, processing, bandwidth, and active user accounts). Control and optimize. Resource usage can be monitored, controlled, and reported, providing transparency for both providers and consumers of the services utilized.

The service model is as follows:

Software as a Service (SaaS): The functionality offered to consumers is to use a provider's applications running on a cloud infrastructure. The application is accessible from a variety of client devices through a thin client interface such as a web browser (eg, web-based email). Consumers do not manage or control the underlying cloud infrastructure, including networks, servers, operating systems, storage, or even individual application functionality, with the exception of limited user-specific application configuration settings. is possible.

Platform as a Service (PaaS): The functionality provided to consumers is provided through consumer-written or off-the-shelf applications created using programming languages and tools supported by the provider. Deploy to cloud infrastructure. Consumers do not manage or control the underlying cloud infrastructure, including networks, servers, operating systems, or storage, but they do have control over applications that host deployed applications and, in some cases, environment configurations. have

Infrastructure as a Service (IaaS): The functionality provided to consumers by providing processing, storage, networking, and other basic infrastructure that allows consumers to deploy and run arbitrary software. Provisioning computing resources, which may include operating systems and applications. Consumers do not manage or control the underlying cloud infrastructure, but do have limited control over operating systems, storage, deployed applications, and in some cases selected networking components (e.g., host firewalls). control.

The deployment model is as follows:

Private cloud: Cloud infrastructure is operated exclusively for one organization. It may be managed by the organization or a third party and may exist on-premises or off-premises.

Community Cloud: A cloud infrastructure is shared by several organizations and supports a specific community with shared issues (e.g., missions, security requirements, policies, and compliance concerns). It may be managed by the organization or a third party and may exist on-premises or off-premises.

Public Cloud: Cloud infrastructure is made available to the general public or large industry groups and is owned by organizations that sell cloud services.

Hybrid Cloud: A cloud infrastructure is a combination of two or more clouds (private, community, or public) that remain a unique entity, but have a standardized structure that allows data and application portability. , or coupled by proprietary techniques (e.g., cloud bursting for load balancing across clouds).

Cloud computing environments are service-oriented with a focus on statelessness, low coupling, modularity, and semantic interoperability. At the heart of cloud computing is an infrastructure that includes a network of interconnected nodes.

Referring now to FIG. 4, an exemplary cloud computing environment 50 is depicted. As shown, cloud computing environment 50 includes a cloud computing environment such as, for example, a personal digital assistant (PDA) or cell phone 54A, a desktop computer 54B, a laptop computer 54C, and/or a vehicle computer system 54N. It includes one or more cloud computing nodes 10 that can communicate with local computing devices used by consumers. Nodes 10 can communicate with each other. These may be grouped physically or virtually in one or more networks, such as private, community, public, or hybrid clouds, or a combination thereof, as described herein above (Fig. (not shown). This allows cloud computing environment 50 to provide infrastructure, platforms, and/or software as a service that does not require cloud consumers to maintain resources on local computing devices. The types of computing devices 54A-N shown in FIG. It will be appreciated that any type of computerized device may be communicated with an addressable connection (e.g., using a web browser) or both.

Referring now to FIG. 5, a set of functional abstraction layers provided by cloud computing environment 50 (FIG. 4) is illustrated. It should first be understood that the components, layers, and functionality illustrated in FIG. 5 are intended to be merely exemplary, and embodiments of the invention are not limited thereto. As depicted, the following layers and corresponding functionality are provided:

Hardware and software layer 60 includes hardware and software components. Examples of hardware components include: mainframes 61, RISC (Reduced Instruction Set Computer) architecture-based servers 62, servers 63, blade servers 64, storage devices 65, and network and networking Component 66. In some embodiments, software components include network application server software 67 and database software 68.

Virtualization layer 70 provides an abstraction layer on which the following examples of virtual entities may be provided: virtual servers 71, virtual storage 72, virtual networks 73 including virtual private networks, virtual applications and operating systems 74, and Virtual client 75.

In one example, management layer 80 may provide the functionality described below. Resource provisioning 81 provides for dynamic procurement of computing resources and other resources utilized to perform tasks within a cloud computing environment. Metering and billing 82 provides cost tracking as resources are utilized within a cloud computing environment and charging or billing for the consumption of these resources. In one example, these resources may include application software licenses. Security provides verification of identity for cloud consumers and tasks, and protection for data and other resources. User portal 83 provides access to the cloud computing environment to consumers and system administrators. Service level management 84 provides allocation and management of cloud computing resources so that required service levels are met. Service level agreement (SLA) planning and execution 85 provides for advance agreement on anticipated future requirements for cloud computing resources and procurement of cloud computing resources according to the SLA.

Workload layer 90 provides an example of the functionality with which a cloud computing environment can be utilized. Examples of workloads and functions that can come from this layer include: mapping and navigation 91, software development and lifecycle management 92, virtual classroom instructional delivery 93, data analysis processing 94, transaction processing 95, and data processing. Governance and XAI Program 175.

The invention may be a system, method, and/or computer program product at every possible level of technical detail of integration. The invention can be advantageously implemented in any system that processes instruction streams, either singly or in parallel. A computer program product may include a computer readable storage medium having computer readable program instructions for causing a processor to perform aspects of the invention.

A computer-readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any suitable combination of the foregoing. . A non-exhaustive list of more specific examples of computer-readable storage media include: portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), Erasable Programmable Read Only Memory (EPROM or Flash Memory), Static Random Access Memory (SRAM), Portable Compact Disk Read Only Memory (CD-ROM), Digital Versatile Disk (DVD), Memory Mechanically encoded devices such as sticks, floppy disks, punched cards or grooved structures with recorded instructions, and any suitable combinations of the foregoing. As used herein, computer-readable storage medium or computer-readable storage device refers to radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., fiber optic cables). It should not be construed as a transient signal itself, such as a pulse of light passing through a wire) or an electrical signal transmitted over a wire.

The computer-readable program instructions described herein may be transferred from a computer-readable storage medium to a separate computing/processing device or over a network, such as the Internet, a local area network, a wide area network, or a wireless network or combinations thereof. can be downloaded to an external computer or external storage device, such as through a network. The network may include copper transmission cables, optical transmission fibers, wireless transmissions, routers, firewalls, switches, gateway computers or edge servers, or combinations thereof. A network adapter card or network interface of each computing/processing device receives computer readable program instructions from the network and receives computer readable program instructions for storage on a computer readable storage medium within the respective computing/processing device. Transfer instructions.

Computer-readable program instructions for carrying out the operations of the present invention include assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state configuration data, configuration data for integrated circuits, or written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk(R), C++, and procedural programming languages such as the "C" programming language or similar programming languages; It may be either source code or object code. The computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer as a stand-alone software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer. It can be run on the computer or on the server. In the latter scenario, the remote computer can connect to the user's computer over any type of network, including a local area network (LAN) or wide area network (WAN), or the connection can be It may be made to an external computer (eg, via the Internet using an Internet service provider). In some embodiments, electronic circuits, including, for example, programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are provided with computer readable program instructions to implement aspects of the invention. The state information can be used to execute computer readable program instructions to personalize electronic circuits.

Aspects of the invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer readable program instructions may be executed by a processor of a computer or other programmable data processing device to implement the functions/acts specified in one or more blocks of the flowchart diagrams and/or block diagrams. may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing device to create a machine. These computer-readable program instructions also provide instructions for a computer-readable storage medium having instructions collectively stored thereon to implement aspects of the functions/operations specified in one or more blocks of the flowcharts and/or block diagrams. may be stored on a computer-readable storage medium to instruct a computer, programmable data processing device, or other device, or combination thereof, to function in a particular manner to provide an article of manufacture that includes an article of manufacture.

Computer-readable program instructions also include instructions for execution by a computer, other programmable apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams. , which is loaded onto a computer, other programmable data processing apparatus, or other device to cause a sequence of operational steps to be performed on the computer, other programmable apparatus, or other device to create a computer-implemented process. It's okay.

The flowcharts and block diagrams in the drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products in accordance with various embodiments of the invention. In this regard, each block of the flowchart or block diagram may represent a module, segment, or portion of instructions, including one or more executable instructions for implementing the specified logical functions. . In some alternative implementations, the functions shown in the blocks may occur out of the order shown in the figures. For example, two blocks shown in succession may actually be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending on the functionality involved. Each block in the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, perform the functions or acts specified, or combine special purpose hardware and computer instructions. It should also be noted that the execution may be implemented by a special purpose hardware-based system.

References herein to "an embodiment," "embodiment," "illustrative embodiment," etc. refer to the embodiment described as including a particular feature, structure, or characteristic, but not necessarily all implementations. Indicates that a form may not include that particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Furthermore, when a particular feature, structure, or characteristic is described in the context of one embodiment, such feature, structure, or characteristic is also discussed in the context of other embodiments, whether or not explicitly described. It is suggested that influencing the structure or properties is within the knowledge of those skilled in the art.

The terminology used herein merely describes particular embodiments and is not intended to limit the invention. As used herein, the singular forms "a," "an," and "the" include the plural unless the context clearly dictates otherwise. It is intended as such. The terms "comprise" and/or "comprising" as used herein refer to the presence of a stated feature, integer, step, act, element, or component or combination thereof. It will be further understood that specifying does not exclude the presence or addition of one or more other features, integers, steps, acts, elements, components or groups or combinations thereof.

Descriptions of various embodiments of the invention have been presented for purposes of illustration and are not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope of the invention. The terminology used herein is used to best explain the principles of the embodiments, their practical application or technical improvements over technology found in the marketplace, or to explain the principles of the embodiments, practical applications, or technical improvements over technology found in the marketplace, or as otherwise known by others of ordinary skill in the art. have been chosen to provide an understanding of the embodiments disclosed in the book.

Claims (20)

A computer-implemented method for providing reproducible machine learning (ML) model results, the method comprising:
receiving input data for the ML model by one or more computer processors;
processing the input data using the ML model to produce an initial result by the one or more computer processors;
adding a first block to a distributed ledger by the one or more computer processors, the first block including the input data, the initial result, the ML model data structure, and the ML model data structure; said adding a link to training data for the model, said training data being present in a preceding distributed ledger block;
providing, by the one or more computer processors, an output that includes the initial result. receiving, by the one or more computer processors, a result validation request regarding the input data;
retrieving the input data, the initial result, and the ML model data structure from the first block by the one or more computer processors;
processing the input data using the ML model data structure to produce a verification result by the one or more computer processors;
2. The computer-implemented method of claim 1, further comprising providing the initial result and the verification result as output by the one or more computer processors. 3. The computer-implemented method of claim 2, further comprising providing an ML model description as output by the one or more computer processors. 2. The computer-implemented method of claim 1, further comprising providing an ML model description as an output by the one or more computer processors, and wherein the first block further includes the ML model description. The computer-implemented method of claim 1, wherein the first block further includes an ML model version label. updating, by the one or more computer processors, the ML model using the input data and results to result in a new ML model version;
adding, by the one or more computer processors, a second block to the distributed ledger, the second block including a label associated with the new ML model version, the input data; 2. The computer-implemented method of claim 1, further comprising said adding a link to training data for said initial result, new ML model data structure, and previous version ML model. linking, by the one or more computer processors, the first block and a user associated with the input data through a distributed ledger entry, the distributed ledger entry allowing the user to obtain information about the initial result; 2. The computer-implemented method of claim 1, further comprising the linking, which can require verification. A computer program product for providing reproducible machine learning (ML) model results, the computer program product comprising one or more computer readable storage devices and one or more computer readable storage devices. program instructions collectively stored on the device, the stored program instructions comprising:
program instructions for receiving input data for the ML model;
program instructions for processing the input data using the ML model and producing an initial result;
Program instructions for adding a first block to a distributed ledger, the first block including input data, an initial result, an ML model data structure, and training data for the ML model. program instructions for said adding, the training data being present in a preceding distributed ledger block;
and program instructions for providing output including said initial result. The stored program instructions are
program instructions for receiving a result verification request regarding the input data;
program instructions for retrieving the input data, the initial result, and the ML model data structure from the first block;
program instructions for processing the input data using the ML model and producing a verification result;
9. The computer program product of claim 8, further comprising program instructions for providing the initial results and the verification results as output. The stored program instructions are
10. The computer program product of claim 9, further comprising program instructions for providing an ML model description as output. The stored program instructions are
9. The computer program product of claim 8, further comprising program instructions for providing an ML model description as output, the first block further comprising the ML model description. 9. The computer program product of claim 8, wherein the first block further includes an ML model version label. The stored program instructions are
program instructions for updating the ML model using the input data and results, resulting in a new ML model version; and program instructions for adding a second block to the distributed ledger, wherein the second the program instructions for adding the new ML model version, the input data, the initial results, the new ML model data structure, and a link to training data for the previous version ML model; 9. The computer program product of claim 8, further comprising: The stored program instructions are program instructions for linking the first block and a user associated with the input data through a distributed ledger entry, wherein the distributed ledger entry allows the user to obtain information about the initial result. 9. The computer program product of claim 8, further comprising program instructions for said linking that can require verification. A computer system for providing reproducible machine learning (ML) model results, the computer system comprising:
one or more computer processors;
one or more computer readable storage devices;
program instructions stored in the one or more computer readable storage devices for execution by the one or more computer processors, the stored program instructions comprising:
program instructions for receiving input data for the ML model;
program instructions for processing the input data using the ML model and producing an initial result;
Program instructions for adding a first block to a distributed ledger, the first block including input data, an initial result, an ML model data structure, and training data for the ML model. program instructions for said adding, the training data being present in a preceding distributed ledger block;
and said stored program instructions for providing an output including said initial result. The stored program instructions are
program instructions for receiving a result verification request regarding the input data;
program instructions for retrieving the input data, the initial result, and the ML model data structure from the first block;
program instructions for processing the input data using the ML model and producing a verification result;
16. The computer system of claim 15, further comprising program instructions for providing the initial results and the verification results as output. The stored program instructions are
17. The computer system of claim 16, further comprising program instructions for providing an ML model description as output. The stored program instructions are
16. The computer system of claim 15, further comprising program instructions for providing an ML model description as output, the first block further comprising the ML model description. The stored program instructions are
program instructions for updating the ML model using the input data and results, resulting in a new ML model version; and program instructions for adding a second block to the distributed ledger, wherein the second the program instructions for adding the new ML model version, the input data, the initial results, the new ML model data structure, and a link to training data for the previous version ML model; 16. The computer system of claim 15, further comprising: The stored program instructions are program instructions for linking the first block and a user associated with the input data through a distributed ledger entry, wherein the distributed ledger entry allows the user to obtain information about the initial result. 16. The computer system of claim 15, further comprising program instructions for said linking that can require verification.

Images